NBER WORKING PAPER SERIES FINANCIAL DEPENDENCE AND INNOVATION: THE CASE OF PUBLIC VERSUS PRIVATE FIRMS. Viral V. Acharya Zhaoxia Xu

Transcription

1 NBER WORKING PAPER SERIES FINANCIAL DEPENDENCE AND INNOVATION: THE CASE OF PUBLIC VERSUS PRIVATE FIRMS Viral V. Acharya Zhaoxia Xu Working Paper NATIONAL BUREAU OF ECONOMIC RESEARCH 1050 Massachusetts Avenue Cambridge, MA December 2013 The views expressed herein are those of the authors and do not necessarily reflect the views of the National Bureau of Economic Research. NBER working papers are circulated for discussion and comment purposes. They have not been peerreviewed or been subject to the review by the NBER Board of Directors that accompanies official NBER publications by Viral V. Acharya and Zhaoxia Xu. All rights reserved. Short sections of text, not to exceed two paragraphs, may be quoted without explicit permission provided that full credit, including notice, is given to the source.

2 Financial Dependence and Innovation: The Case of Public versus Private Firms Viral V. Acharya and Zhaoxia Xu NBER Working Paper No December 2013 JEL No. G31,G32,O16,O30 ABSTRACT This paper examines the relationship between innovation and firms' dependence on external capital by analyzing the innovation activities of privately-held and publicly-traded firms. We find that public firms in external finance dependent industries generate patents of higher quantity, quality, and novelty compared to their private counterparts, while public firms in internal finance dependent industries do not have a significantly better innovation profile than matched private firms. The results are robust to various empirical strategies that address selection bias. The findings suggest that public listing is beneficial to the innovation of firms in industries with a greater need for external capital. Viral V. Acharya Stern School of Business New York University 44 West 4th Street, Suite 9-84 New York, NY and CEPR and also NBER vacharya@stern.nyu.edu Zhaoxia Xu Department of Finance and Risk Engineering New York University 6 MetroTech Center, New York, NY zhaoxiaxu@nyu.edu

3 1 Introduction While innovation is crucial for businesses to gain strategic advantage over competitors, financing innovation tends to be difficult because of uncertainty and information asymmetry associated with innovative activities (Hall and Lerner (2010)). Firms with innovative opportunities often lack capital. Stock markets can provide various benefits as a source of external capital by reducing asymmetric information, lowering the cost of capital, as well as enabling innovation in firms (Rajan (2012)). Given the increasing dependence of young firms on public equity to finance their R&D (Brown et al. (2009)), understanding the relation between innovation and a firm s financial dependence is a vital but under-explored research question. We fill this gap in the literature by investigating how innovation depends on the access to stock market financing and the need for external capital. We use a firm s public listing status to capture the access to stock markets and investigate its impact on innovation. While firms can gain an access to a large pool of low cost capital by trading on exchanges, they also face the pressure from myopic investors to generate short-term profits (Stein (1989)). Therefore, we expect that the effect of public listing on innovation will depend on the trade-off between the benefits and costs associated with listing on stock markets, which vary across firms with different degrees of dependence on external finance. By analyzing the innovation activities of a large sample of private and public firms between 1994 and 2004, we observe that public firms on average have patents of higher quantity, quality, and novelty than private firms. After considering the need for external 1

4 finance, we find that only public firms in external finance dependent (EFD) industries have a significantly better innovation profile than private firms, not firms in internal finance dependent (IFD) industries. Industries that rely on external (internal) finance for their investments are considered as EFD (IFD) industries. To understand why public firms in industries with a greater need of external capital perform better in innovation, we explore four potential explanations. One potential explanation could be that public listing relaxes the financial constraints faced by those firms. If this is the case, one would expect that firms with a higher propensity to innovate will benefit more from obtaining access to stock market financing. Consistent with this conjecture, we find that firms in more innovation intensive industries with a greater need for external capital are more likely to go public compared to firms in less innovation intensive industries and they are more innovative when having access to public equity capital compared to firms without such access. The observed difference in innovation may also be driven by the variation in firms ability to use R&D to generate patents. To explore this second possibility, we test whether public and private firms differ in their innovation efficiency measured as the natural logarithm of one plus the number of patents per dollar R&D investment. We find a higher innovation efficiency for public firms in industries dependent on external finance, but no significant difference for private and public firms in IFD industries. Third, a part of the literature has argued that public firms are prone to agency problems given the separation of ownership and control. Under the pressure of myopic investors, managers have incentives to pursue short-term performance (Stein (1989), 2

5 Bolton et al. (2006)). 1 In light of the short-termism model, we investigate public firms real earnings management activities in relation to their degree of external finance dependence and innovation. We find that more innovative public firms in EFD industries engage less in earnings management through their alteration of their real activities. To the extent that real earnings management represents firms myopic behavior, innovative firms with a greater need for external capital appear less likely to boost short-term earnings at the expense of long-term values. Fourth, the better innovation profile of public firms in EFD industries may be a result of patent acquisitions outside firm boundaries. Recent studies provide evidence that public firms have incentives to purchase patents and new technologies through mergers and acquisitions (Bena and Li (2013), Seru (2013)). Sevilir and Tian (2013) show that acquiring innovation can enhance the innovative output of the acquirers. Since the access to stock markets can provide the capital needed for patent purchase, this acquisition-based explanation is actually consistent with the view that public listing provides financing benefits for innovation. Overall, our results suggest that financing benefits coupled with innovation efficiency and innovative firms lower incentives to behave myopically help to explain the difference in the innovation of public and private firms in EFD industries. Perhaps the biggest challenge of our empirical design is the concern that a firm s decision to gain access to stock markets may be an endogenous choice driven by other 1 In September 2009, the Aspen Institute along with 28 leaders including John Bogle and Warren Buffett called for an end of value-destroying short-termism in U.S. financial markets and an establishment of public policies that encourage long-term value creation (Aspen Institute (2009)). 3

6 observed and unobserved factors. To overcome this selection bias, we adopt several identification strategies enabled by our large panel dataset of U.S. private and public firms. Our fixed effects estimation explicitly controls for observable time-series and crosssectional variables that are related to innovation and the decision of going public. We then employ an econometric method to directly adjust for selection bias from unobservables. Specifically, we estimate the treatment effect model using an inverse Mills ratio to explicitly correct for selection bias. 2 Furthermore, we adopt several quasi-experimental designs to alleviate the concern about the non-randomness of public and private firms. The first quasi-experiment applies the propensity score matching method to identify a sample of firms that transition from private to public (treatment group) and a sample of similar firms that remain private (control group). The difference-in-differences approach is then used to isolate the treatment effect by differencing out the influence of crosssectional heterogeneity or common time trends on the innovation activities of the treated and the controlled groups. Identification of this approach relies on the assumption that the closely matched private firms act as a counterfactual for how the transition firms would have performed without going public. We observe a positive treatment effect in the patent portfolios for firms in EFD industries, while the effect is mostly insignificant for firms in IFD industries. To ease the concern that a firm may go public at a specific stage of its life cycle, we adopt a second quasi-experiment, which we construct two groups of firms: a treatment 2 We also estimate an instrumental variable model using the percentage of public firms in the industry in a given year as an instrument for being public. A firm is more likely to go public as their peers in the same industry sell their shares publicly (Scharfstein and Stein (1990)), but its innovation activities are unlikely to be affected by the percentage of publicly-traded firms in the same sector other than through the publicly listing channel. The results are reported in the Appendix Table A.1. 4

7 group consisting of firms that eventually completed the initial public offering (IPO) after the withdrawal of the initial registration statement with Securities and Exchange Commission (SEC) and a control group of firms that ultimately did not go public after the initial withdrawal. 3 Applying the triple differences approach in a multivariate framework, we find an increase in the quantity and originality of patents for firms that successfully transition from private to public. Furthermore, this improvement in patent portfolios is concentrated in firms in EFD industries. The triple differences approach relies on the assumption that the average outcome variables follow a parallel trend over the pre-treatment period. The validation of this parallel trend assumption is verified in our graphical test. Figure 1 shows that the trends in patents for both treatment and control groups are similar during the pre-withdrawn and pre-ipo eras, while the number of patents in the treatment group increases significantly after an IPO. Our multivariate test also confirms that there is no systematic difference in the trend of patents between the treatment and control group during the pre-treatment era. The third quasi-experiment involves a fuzzy regression discontinuity design exploiting the discontinuous nature of NASDAQ listing requirements for assets. The NASDAQ requires that a listed firm have a minimum number of net tangible assets. 4 Identification of this design relies on the assumption that observations close to the discontinuity 3 The process of going public in the U.S. requires filing security registration documents with the SEC. After the registration, the filers still have the option to withdraw their offering before issue. Withdrawals of registered IPOs are not uncommon. Dunbar and Foerster (2008) examine the period and document that about 20% of firms withdrew their IPO filings and 9% of the withdrawn firms successfully complete the process later. 4 See Section 5.4 for details of the requirement. 5

8 threshold are similar. We first conduct a graphic analysis of the relationship between patent portfolios and the forcing variable (normalized net tangible assets in the IPO year) around the threshold. Figure 2 shows that firms with net tangible assets above the cutoff have a better patent portfolio than firms with net tangible assets below the cutoff. Moreover, the placebo analysis that uses normalized net tangible assets in a random year as the forcing variable exhibits no jump in patent portfolios at the threshold (Figure 3). Our formal fuzzy regression discontinuity estimations indicate that IPO firms listed on the NASDAQ have a relatively stronger innovation profile compared to private firms with net tangible assets very close to the minimum listing requirements of the NASDAQ. Our study is related to the nascent literature on identifying various economic factors driving firm innovation. The literature shows that innovation is affected by the development of financial markets (Amore et al. (2013), Chava et al. (2013), Hsu et al. (2013)), legal system (Brown et al. (2013)), bankruptcy laws (Acharya and Subramanian (2009)), labor laws (Acharya et al. (2013)), competition (Aghion et al. (2005)), investors tolerance for failure (Tian and Wang (2012)), institutional ownership (Aghion et al. (2013)), and private equity (Lerner et al. (2011)) 5. Differing from previous work focusing on public firms, we analyze a large sample of private and public firms and find that the innovation capacity of firms in external finance dependent industries is influenced by 5 Lerner et al. (2011) find no evidence that private equity sacrifices innovation to boost short-term performance using a sample of 472 leveraged buyout (LBO) transactions during In a similar spirit, we identify firms that experienced LBOs based on our sample ( ) and explore changes in innovation of these firms in comparison with the matched public firms based on firm characteristics. Our unreported results from propensity score matching coupled with difference-in-differences estimations show no significant difference in changes in innovation during the transition between the LBO firms and the controlled public firms. 6

9 access to stock market financing. This paper also adds new evidence to the recent surge of debate on the trade-off between public listing and staying private and its influence on firms real activities. On the one hand, the benefits of an easier access to cheaper capital allow public firms to conduct more mergers and acquisitions (Maksimovic et al. (2012)), to raise more equity capital (Brav (2009)), and to pay more dividends (Michaely and Roberts (2012)) than private firms. Public firms can take better advantage of growth opportunities and are more responsive to changes in investment opportunities than their private counterparts (Mortal and Reisel (2012)). On the other hand, the agency conflicts resulting from divergent interests between managers and investors at public firms distort their cash holdings (Gao et al. (2013)), investments (Asker et al. (2011)), and innovation (Bernstein (2012)). 6 Our findings suggest that the lower cost of capital associated with public listing are important for innovation of firms with large capital needs, while the financing benefits of stock markets are weaker for innovation of firms in internal finance dependent industries. The rest of the paper is organized as follows. We develop hypotheses in Section 2. In Section 3, we describe the data, innovation, and external finance dependence measures. Section 4 presents differences in innovation of private and public firms. In Section 5, 6 Differing from Bernstein (2012) who focuses only on the innovation activities of IPO firms, we investigate the innovation of public and private firms in general and link it to their financial dependence. Among the battery of identification strategies that intend to address the endogeneity concern, one of our analyses involves a subsample of firms that experience the IPO transition. Distinct from Bernstein (2012) s comparison of patents of successful IPO firms with IPO withdrawn firms, we investigate a group of firms with shared experience, that is, firms that eventually completed the IPO process following the withdrawal of their initial filings and firms that ultimately did not go public after the withdrawal. Our results suggest that the effect of public listing on innovation depends on firms need on external capital. 7

10 we exploit several quasi-experimental designs to isolate the treatment effects. Section 6 discusses the potential explanations for the observed difference in innovation of private and public firms. We conclude in Section 7. 2 Theoretical Motivation and Empirical Hypothesis The theoretical literature presents two opposing views on the impact of stock markets on innovation. One view focuses on the myopic nature of stock markets and/or managers. These models of short-termism argue that stock markets tend to be obsessed with short-term earnings and such myopia could induce public firms to invest sub-optimally (Stein (1989); Bebchuk and Stole (1993)). With their compensation linked to stock performance, managers of public firms have incentives to sacrifice long-term investments in order to boost short-term stock returns. Innovation typically requires a substantial amount of investments for a long period of time and the probability of success is highly uncertain. Holmstrom (1989) and Acharya et al. (2013) suggests that managers, under the pressure to establish a good performance record in capital markets, have few incentives to undertake long-term investments such as innovation. Moreover, with the assumption of observable cash flows and no tolerance for failures in public companies, Ferreira et al. (2012) develop a model to demonstrate that managers of public companies are rationally biased against innovative projects, which usually have a higher failure rate. An implication of these models is that stock markets hinder managers from investing in innovation. 8

11 The other view focuses on the financing advantages of stock markets for innovation. First, stock markets are an important source of financing for innovation. Allen and Gale (1999) model indicates that public equity markets, which allow investors with diversified opinions to participate, enable the financing of innovative projects with uncertain probabilities of success. As illustrated in the model of Rajan (2012), the ability to secure captial alters the innovative nature of firms. Equity markets play an essential role in providing the capital and incentives that an entrepreneur needs to innovate, transform, create enterprise, and generate profits. He argues that firms with an easier access to equity capital are more likely to conduct capital-intensive fundamental innovation. Second, the literature suggests that equity is preferable to debt in financing innovative projects. Hall and Lerner (2010) suggest that intangible assets and knowledge created by innovation are difficult to quantify as collateral for debt financing. The uncertainty and volatile return of innovative projects also make them unattractive to many creditors (Stigliz (1985)). Moreover, Rajan (2012) points out that the possibility of losing the critical asset to creditors in the event of project failure discourages entrepreneurs to innovate. In contrast, equity capital is a favorable way to finance innovation since it allows investors to share upside returns and does not require collateral. Third, the listing in a stock market lowers the cost of capital as investors portfolios become more liquid and diversified (Pagano et al. (1998); Benninga et al. (2005)). It also helps to lower borrowing costs because of the reduced asymmetry of information (Schenone (2010)) and increased lender competition (Saunders and Steffen (2011)). Given the contrasting predictions of the two streams of research, it becomes an empir- 9

12 ical question as to how stock markets actually affect innovation. With the implications of theoretical models in mind, we conjecture that the impact of listing in stock markets on innovation varies with the degrees of external finance dependence. Rajan and Zingales (1998) argue that industries differ in their demand for external financing due to the differences in the scale of the initial and continuing investments, the incubation period, and the payback period. For firms with excess cash flows over their investment needs, the infusion of public equity should not affect the marginal cost of capital and therefore may not increase their innovation. With the exposure to stock market short-termism, going public might even potentially stifle the innovative activities of those firms. However, for firms with insufficient internal cash flows for their investments, the additional capital raised from stock markets could relax their financial constraints and facilitate innovation. Consequently, stock markets should matter more for firms in industries with a greater need for external funds. Considering the differential needs for external capital, we hypothesize that public listing should promote the innovation of firms in industries dependent more on external finance. 3 Data and Innovation Measure 3.1 Data To measure innovation activities, we collect firm-year patent counts and patent citations data from the latest edition of the National Bureau of Economic Research (NBER) Patent Citation database. The database contains information on every patent granted by 10

13 the United States Patent and Trademark Office (USPTO) from 1976 to 2006, including patent assignee names, the number of citations received by each patent, a patent s application year, a patent s grant year, and the technology class of the patent, among other items. The financial data on U.S. private and public firms are obtained from S&P Capital IQ for the The sample stops in 2004 because the average time lag between patent application date and grant date is two to three years (Hall et al. (2001)). 7 S&P Capital IQ categorizes a firm as public or private based on its most recent status. For example, Google Inc. is classified as public in 2002 although it went public in We reclassify a firm s private (or public) status with IPO date from Compustat, Thomson One, Jay Ritter s IPO database, the first trading date information from CRSP, and delisting date information from Compustat. Financial institutions and utilities (SIC code and ) and firms with no SIC codes are excluded. We require non-missing data on total assets and non-negative value on total revenue. Firm-years with total assets less than $5 million USD are excluded. Cash, leverage, capital expenditure ratios, and R&D ratios are winsorized at 1% and 99% to avoid the effect of outliers. We merge financial data with the patent database by GVKEY and by company names when GVKEY is unavailable. We manually check the names to ensure the accuracy of the match. In cases where the names are not exactly identical, we conduct internet searches and include the observation only if we are confident of the match. Following the innovation literature (e.g. Atanassov (2013)), the patent and citation counts are 7 Using a sample period of 1994 to 2003 yields similar results. 11

14 set to zero when no patent and/or citation information is available. Including firm-year observations with no patents alleviates the sample selection concern. The final sample has 2,392 private firms and 8,863 public firms. Previous studies have shown that innovation varies substantially across industries and by firm size (Acs and Audrestsch (1988)). To minimize the differences in industry and size distributions, we identify an industry-and-size-matched sample of private and public firms. Specifically, for each private firm from the beginning of the sample period, we find a public firm closest in size and in the same four-digit SIC industry. 8 The timeseries observations for each match are kept in order to preserve the panel structure of the data. This procedure results 1,717 matched pairs of private and public firms. 3.2 Innovation Measure We use R&D spending to measure innovation input and patent-based metrics to measure innovation output (Hall et al. (2001, 2005)). The first measure of innovation output is the number of patents applied by a firm in a given year. The patent application year is used to construct the measure since the application year is closer to the time of the actual innovation (Griliches (1990)). Patent innovation varies in their technological and economic significance. A simple count of patents may not be able to distinguish breakthrough innovations from incremental technological discoveries (Trajtenberg (1990)). Thus, we use the citation count each patent receives in subsequent years to measure the 8 Closest in size means that two firms have the smallest ratio of their total assets (TA). The ratio of total assets is defined as max(t A private, T A public )/min(t A private, T A public ). Asker et al. (2011) use a similar method to identify firm s closest in size. 12

15 importance of a patent. Citations are patent-specific and are attributed to the applying firm at the time of application, even if the firm later disappears due to acquisition or bankruptcy. Hence, the patent citation count does not suffer survivorship bias. Hall et al. (2005) show that the number of citations is a good measure of the quality of an innovation. However, the patent citation is subject to a truncation bias. This is because citations are received over a long period of time, but we only observe the citations up to Compared to patents created in earlier years, patents created in later years have less time to accumulate citations. Additionally, the citation intensities of patents might vary across different industries. Lerner et al. (2011) suggest that the frequency of patent citations, as well as patents in technologically dynamic industries have increased in recent years. To correct for this time trend in citations, we scale the raw patent citation counts by the average citation counts of all patents applied in the same year and technology class following Hall et al. (2001, 2005). 9 This measure shows the relative citation counts compared to matched patents after controlling for time and technology fixed effects. Innovative projects differ in their novelty. Fundamental research tends to be risky and produce more influential innovations. Following Trajtenberg et al. (1997), we use the originality and generality of patents to measure the novelty of innovation. These two proxies also reflect the degree of risk that firms are bearing in their pursuit of R&D. 9 An alternative way to adjust patent citations for truncation bias is to weight the number of citations with the estimated distribution of citation-lag. That is, each patent citation is adjusted using the citation truncation correction factor estimated from a diffusion model. The weakness of this adjusted citation is that it does not measure the relative importance of the patent compared to similar patents. Using this truncation-bias-adjusted citation yields similar results. 13

16 Originality is computed as the Herfindahl index of cited patents: n i Originality i = 1 where F ij is the ratio of the number of cited patents belonging to class j to the number of patents cited by patent i. The originality of a patent indicates the diversity of the patents cited by that patent. A patent that cites a broader array of technology classes has a higher originality value. j F 2 ij, Similarly, generality is measured as the Herfindahl index of citing patents: n i Generality i = 1 where G ij is the number of patents citing patent i belonging to class j scaled by the number of patents citing patent i. The generality of a patent indicates the diversity of the patents citing that patent. A patent that is cited by a broader array of technology classes has a higher value of generality. j G 2 ij, 3.3 External Finance Dependence Measure Rajan and Zingales (1998) argue that the degree of dependence on external financing varies across different industries. Industries such as biotechnology rely more on external capital, while industries such as tobacco are less external capital dependent. To construct an industry s dependence on external finance, we follow Rajan and Zingales (1998) and first measure a firm s need for external finance in a year as the fraction of capital expenditure not financed through internal cash flow. The time series industry-level external finance dependence is constructed as the median value of the external finance 14

17 needs of all firms in the two-digit SIC code industry in each year. We then measure each industry s external finance index as a percentile ranking of its time series median during An industry with a higher index value of external finance dependence relies more on external capital to finance its investment. 4 Empirical Analysis 4.1 Univariate Analysis In Table 1, we compare firm characteristics and innovation activities of private and public firms in the full sample (Panel A) and the matched sample (Panel B). In the full sample, public firms on average are bigger in size and older compared to private firms. firm. 11 Age is defined as the difference between current year and founding year of a Private firms have more tangible assets and higher sales growth. In terms of cash holdings, private firms hold a lower percentage of their assets as cash (14.66% of total assets), while public firms reserve a higher percentage of cash (18.89% of total assets). The average return on assets (ROA) of private firms is lower than that of public firms. Private firms have a capital expenditure ratio of 7.20% relative to total assets, while public firms have a ratio of 6.31%. As for innovation activities, Panel A of Table 1 shows that public firms have a slightly lower R&D ratio, defined as R&D expenses as a ratio of total assets, than private firms. 10 Hsu et al. (2013) use a similar approach to measure an industry s dependence on external finance. 11 To compute firm age, we cross-check the founding year data in Capital IQ and Jay Ritter IPO databases to ensure accuracy. 15

18 The ratio of R&D expenditure to total assets is 5.48% for private firms, while the ratio is 4.93% for public firms. In terms of the outcome of investments in innovation, private companies on average have significantly fewer patents compared to public firms (1 vs. 7). The patents applied by public firms are on average of better quality than those of private companies as measured by the truncation bias adjusted citations. The patents of public companies receive more citations compared to those of private companies (0.32 vs. 0.18). The difference in the average number of citations to the patents of private and public firms is statistically significant. Public firms also tend to produce more original patents with wider applications. Similar differences between private and public firms are observed in the matched sample, with a few exceptions. Panel B of Table 1 shows that the matched private and public firms are similar in size after we match firms on size and industry. Public firms have fewer tangible assets, lower sales growth, fewer tangible assets, more cash, lower ROA, and lower capital expenditure ratios than otherwise similar private firms. For the size-and-industry matched sample, public firms on average have a higher R&D ratio. The patent profile of matched public firms is better than their private counterparts. For example, the average number of patents generated by public firms is 2, while it is fewer than 1 for matched private firms. 4.2 Multivariate Analysis The univariate analysis indicates that public firms on average outperform private firms when it comes to their innovation activities. However, the difference in innovation 16

19 outcome between private and public firms may be confounded by the difference in firm characteristics. To control for the distinctness in observable firm attributes and the influences of industry characteristics and time on innovation, we estimate the following panel data model: Y ikt = α + βp ublic i + γx ikt + η k + ζ t + ε ikt, (1) where Y ikt measure innovation activities. The measures include R&D ratio, number of patents, truncation bias adjusted citations, originality, and generality. P ublic i is a dummy variable equal to one for public firms and zero for private firms; X ikt is a set of characteristic variables that affect a firm s innovation activities, including ln(sales) (log of total revenue), T angible (tangible assets scaled by total assets), Cash (total cash scaled by total assets), Age (the difference between current year and founding year); Capex (capital expenditures scaled by total assets), S.Growth (the first difference of the natural logarithm of total revenue), ROA (EBITDA divided by total assets); η k control for industry effects based on two-digit SIC codes; and ζ t control for year fixed effects. The coefficient β estimates the effect of public listing on innovation while the confounding variables are controlled. Since the full sample and the industry-and-size matched sample yield similar results, we report the main results based on the matched sample. In Panel A of Table 2, the first specification has R&D ratio as the dependent variable. The coefficient on the dummy variable P ublic is positive, indicating that public firms spend more on R&D than private firms once the confounding effects have been controlled. R&D ratio of public firms is 0.48% higher than matched private firms. With regard to the outcome of investments in 17

20 innovation, there is a significant difference between the two types of firms. The estimated coefficients on P ublic are positive and significant in all specifications. Public firms on average have one more patents than private firms. The patents of public firms are also more influential in terms of citations compared to those of private firms. The originality and generality of the patents developed by public firms are also higher than those by private firms. As for control variables, we observe that larger firms tend to have a higher R&D ratio, produce more patents, receive more citations to their patents, and have more novel innovation. Firms with more tangible assets produce more patents that have a broader impact. The coefficients on Cash are positive and significant, which suggests that firms with more cash are more innovative. The incentives to invest in innovation may vary among firms during different stages of their lifecycles. We use the age variable to control for a firm s lifecycle effects. Mature firms tend to have lower R&D spending as a percentage of total assets. Regarding innovation outcome, there is no significant difference between older and younger firms in terms of patent quantity and citations. However, patents produced by older firms are more novel. The coefficients on Capex are positive but insignificant in general. The coefficients on ROA are negative, while those on sales growth are mixed. 4.3 Treatment Effect Model Estimation The panel data estimations provide suggestive evidence that the public listing status of a firm is associated with its innovative ability. Clearly the decision of being public or 18

21 staying private is not random. The effect of treatment (being public) may differ across firms and may affect the probability of firms going public. To establish causality, we need to control for unobservables that could drive both innovation and the decisions to go public. To address the potential endogeneity of the treatment dummy, we estimate the treatment effect model that explicitly corrects for selection bias using the inverse Mills ratio. The treatment effect model includes two equations. The first one is the outcome equation (equation (1)) with the dummy variable P ublic indicating the treatment condition (i.e., being public). The coefficient β denotes the average treatment effect: AT E = E(Y i P ublic = 1) E(Y i P ublic = 0). The second one is the selection equation: 1 if P ublic i > 0 P ublic i = P ublic i = π + δz i + υ i (2) 0 if P ublic i 0 where Z is a set of firm characteristic variables that affect a firm s decision to go public. The treatment model is estimated with a two-step approach. The first step estimates the probability of being public from the probit model in equation (2). The second-step includes the inverse Mills ratio (Mills) to equation (1) in order to adjust for the selfselection bias. We report the first step of the estimation in the first column of Table A.2. The results for the second step of the estimation are reported on Panel B of Table 2. The negative coefficient on the inverse Mills ratio indicates that the covariance between the error terms in the selection and outcome equations is negative. Firms are more likely to choose go public when the impact on innovation is smaller. The coefficients on the P ublic dummy are all positive and significant. After correcting for selection bias, public 19

22 firms still appear to spend more on R&D, get more patents, and have higher quality and more novel innovation. Public firms R&D to total assets ratio is 1.24% higher than the size-and-industry matched private firms. Public firms on average produce three more patents per year compared to their private counterparts. 4.4 External Finance Dependence and Innovation To investigate the relationship between innovation and a firm s access to stock market financing conditional on its need for external finance, we classify firms into external finance dependent and internal finance dependent industries. We regard industries with a positive value of the external finance dependence measure as external finance dependent, while those with a negative value as internal finance dependent. We first compare the characteristics and innovation of private and public firms in external and internal finance dependent industries. Table 3 shows that the differences in characteristics between private and public firms are similar among industries with differential levels of dependence on external finance. Regarding innovation activities, public firms produce significantly more patents than private firms and their patents are more important and of better quality too. The differences between private and public firms are larger in EFD industries than in IFD industries. The average difference in patent is 1.54 for public and private firms in EFD industries, while the difference is 0.23 for those in IFD industries. We then estimate the treatment effect model separately for firms in EFD and IFD industries. Table 4 shows that the coefficients on the dummy variable P ublic are pos- 20

23 itive and significant for firms in EFD industries, but are insignificant for firms in IFD industries. 12 The result suggests that being publicly listed has a stronger impact on innovation in industries with a greater need for external capital. For example, public firms on average have about 4 more patents than private firms in EFD industries, while the difference between public and private firms is negative and insignificant in industries dependent less on external capital. The patents of public firms in the EFD industries are also more important. Additionally, the differences in the originality and generality of patents produced by public and private firms are only significant in EFD industries. To test whether the impact of public listing on innovation is significantly different between EFD and IFD industries, we include several interaction terms to the second step of the treatment effect model. The estimated model is as following: Y ikt = α+βp ublic i +δef D ik +θp ublic i EF D ik +γx ikt 1 +λx ikt 1 EF D ik +φmills i +ε ikt, (3) where EF D ik is the industry external finance index. Panel C of Table 4 reports the coefficients on θ. The coefficients are positive and significant, indicating that the impact on innovation of being publicly listed is stronger in EFD industries than in IFD industries. Overall, the results are consistent with the view that having a public listing status positively affects the innovation of firms with a greater need for external capital. 12 To ease the concern about the imbalance in the number of firms in EFD and IFD industries, we divide firms in external finance dependent industries into tertiles and estimate the treatment effect model using firms in the top tertile. The results are reported in Table A.3. We still observe that public firms in external finance dependent industries have relatively better innovation profiles than private firms and the difference is statistically significant. 21

24 5 Quasi-Experiments The estimations so far are based on the treatment effect model, which directly controls for selection bias through an inverse Mills ratio. To further ease the concern about the non-randomness of public and private firms, we explore three quasi-experimental designs: (1) the propensity score matching (PSM) combined with the difference-indifferences (DD) approach that compares firms transitioning from private to public with those remain private, (2) the triple differences (DDD) approach investigating firms that experienced withdrawal of an IPO, and (3) a fuzzy regression discontinuity approach investigating discontinuity in the probability of going public as a function of NASDAQ listing requirement for net tangible assets. These quasi-experiments are used to isolate the causal effect of public listing on innovation. 5.1 Difference-in-Differences The first quasi-experiment uses the DD approach involving two groups: a treatment group consisting of firms transitioning from private to public during the sample period and a control group including firms that remain private. To estimate the treatment effect, we compare the changes in the outcome variables of the treatment group (before and after the implementation of the treatment) with those of the control group. Following the suggestion of Blundell and Dias (2000), we combine the PSM with the DD approach. To investigate the dynamics, we require firms to have at least four consecutive years of data and require IPO firms to have data at least two years before 22

25 and one year after the IPO. We use the PSM method to match the IPO firms and private firms by the propensity scores of being public from the logit regression based on their total assets, capital expenditure, ROA, and leverage. 13 The matched firms are required to operate in the same industry. The sample used for the logit regression includes 961 IPO firms and 695 private firms. We use the year that an IPO firm goes public as the fictitious IPO year for its matched private firm. The matched sample consists of 370 pairs of private and IPO firms; 318 pairs are in external finance dependent industries. After obtaining the closely matched treatment and control groups, we apply the DD approach to difference out the cross-sectional heterogeneity or common time trend that affects both groups of firms. Panel A of Table 5 presents the results from the DD analysis for firms in EFD industries. We compute the DD estimator as the difference of changes in the average patent portfolios of the treatment and control groups around the IPO. For external finance dependent industries, firms that transition from private to public experience an increase in the number of patents, and patent citations, as well as the originality of the patents, while firms that remain private experience a marginal decrease in patents. R&D as a percentage of total assets declines slightly after firms go public, although the dollar amount of spending on innovation development increases. The DD for the treatment and the control groups in EFD industries are statistically significant, except for generality (Panel A). However, the DD for patent portfolios of the treatment and control groups in IFD industries are generally insignificant (Panel B). To the extent that the innovation activities of the private firms represent the counterfactual 13 We use propensity score matching with no replacement and a caliper of 0.25 standard deviation. 23

26 scenario if the IPO firms did not go public, the results provide no evidence that going public impairs a firm s ability to innovate, especially for firms in EFD industries. 5.2 Triple Differences A potential concern with the first quasi-experiment is that the treatment effect may be confounded by a firm s choice of the timing of its IPO. Therefore, we explore the second quasi-experiment which involve firms that withdrew their IPO registrations for reasons unrelated to innovation and adopt a DDD approach. The treatment group includes firms that eventually completed the IPO after the initial withdrawal (success sample). The control group comprises of firms that ultimately failed to go public (withdrawn sample). The withdrawn sample can act as a counterfactual for how the success sample would have performed if they failed to go public. We focus on firms that experienced withdrawal of an initial registration statement for two reasons. First, it eases the concern that a comparison of innovation dynamics of IPO firms around the transition with the matched private firms may simply reflect the difference in the lifecycles of those firms. Second, it minimizes the concern that a comparison of the innovation activities of IPO firms without the experience of IPO filing withdrawal with those of withdrawn firms may be confounded by endogeneity of the decision to withdraw. We identify firms that withdrew their initial registrations from S&P Capital IQ and Thomson One equity issuance databases and apply the DD and DDD estimations. Our identification strategy compares innovation activities (1) before and after IPO, (2) across 24

27 the success and withdrawn samples, and (3) across firms in the EFD industries and the IFD industries. The DDD estimating equation is thus: Y ikt = α + βsuccess i + δsuccess i After it + θafter it + δef D ik (4) + θsuccess i EF D ik + κsuccess i After it + ρsuccess i After it EF D ik + γx ikt 1 + λx ikt 1 EF D ik + φmills i + ε ikt, where Y ikt is the measures of innovation activities: R&D, number of patents, truncationbias adjusted citations; Success i is a dummy variable equal to one for firms that completed an IPO after withdrawal of the initial filing and zero for firms that did not complete an IPO after withdrawal of the initial filing; After it is a dummy variable that takes a value of one for post-withdrawn years of withdrawn firms and post-ipo years of successful IPO firms; EF D ik is an industry external finance index; and X ikt 1 is a set of characteristic variables that affect a firm s innovation activities. Table 6 reports the results of DD (Panel A) and DDD (Panel B) estimations. Panel A shows that the coefficients on Success Af ter are insignificant, suggesting that, on average, there is no significant difference in the innovation of successful and withdrawn firms in all industries. In Panel B, we condition our analysis on firms dependence on external capital. The coefficient (δ) represents the differential post-ipo impact between the treatment and control groups in IFD industries. The negative coefficients in all specifications suggest no improvement in the innovation profile of firms in IFD after they complete an IPO. The coefficients on the three-way interactive term (ρ) are significant and positive in the specifications of patent and originality. The positive coefficients 25

28 indicate that external finance dependent firms that eventually went public produce more patents after IPO. The patents of these successful IPO firms are of higher originality than the patents produced before IPO. The coefficients are positive but not significant in the specifications of citations and generality. Overall, the DDD results are consistent with the view that the access to stock markets helps the innovation of firms in a greater need of external capital. 5.3 Parallel Test The key identifying assumption of DDD approach is the parallel trend assumption under which, in absence of treatment, the average outcomes for the treatment and control groups would have the same variation. We perform two diagnostic tests to ensure the parallel trend assumption is satisfied. The first test is a graphic diagnosis. We plot the patent dynamics of the treatment group over the pre-withdrawn, pre-ipo, and post- IPO periods and that of the control group over the pre-withdrawn and post-withdrawn periods. 14 Figure 1 shows that the treatment and control groups follow similar trends in patents during the pre-withdrawn and pre-ipo eras. As a second test to investigate whether or not there is pre-trend in innovation prior to the transition from private to public, we adopt an approach similar to Bertrand and Mullainathan (2003) and Acharya and Subramanian (2009). We use four dummy variables to capture any effects during four separate time periods: before withdrawal of the initial registration statement (P re-w ithdrawn); during the period between the 14 In order to examine changes in patents around the transitions, we require that firms in the treatment group have at least one observation in each of the three periods. 26

29 withdrawn year and the IPO year (P re-ip O); the IPO year and one year after the IPO (After 0,1 ); and two years after the IPO and beyond (After 2+ ). The following model is estimated: Y ikt = α+βp re-w ithdrawn it +δp re-ip O it +θafter 0,1 it +δafter 2+ i +γx ikt 1 +ε ikt. (5) We find that the coefficients on the dummy variables P re-w ithdrawn and P re-ip O are all statistically insignificant (Table 7). There is no evidence of a pre-trend. The coefficients on After 2+ are positive and significant in the specifications of patent and generality, which suggests that innovation begins to increase two years after the completion of an IPO. 5.4 Regression Discontinuity As the third strategy to examine the causal effect of an IPO on innovation, we apply a quasi-experimental fuzzy regression discontinuity (RD) design discussed in Angrist and Lavy (1999) and Hahn et al. (2001). Identification in a fuzzy RD relies on the assumption that observations sufficiently close to the discontinuity threshold (x 0 ) are similar. Fuzzy RD exploits discontinuity in the probability of treatment as a function of the forcing variable (x i ) and uses the discontinuity as an instrumental variable for treatment. 15 In our context, we use the log normalized NASDAQ listing requirement for net tangible assets as the forcing variable x i and exploit discontinuity in the probability of an IPO 15 Sharp regression discontinuity is not suitable for studying public listings because an IPO is not solely determined by the observable listing criteria. The probability of treatment (IPO) is affected by factors other than the forcing variable. Thus, the probability of treatment does not jump from 0 to 1 when the forcing variable crosses the threshold. 27